PHIL 13-1 Philosophical Perspectives, 13, Epistemology, 1999

A PRIORI WARRANT AND NATURALISTIC EPISTEMOLOGY

The Seventh Philosophical Perspectives Lecture

Alvin I. Goldman The University of Arizona

1. Introduction Epistemology has recently witnessed a number of efforts to rehabilitate rationalism, to defend the existence and importance of a priori knowledge or warrant construed as the product of rational insight or apprehension (Bealer 1987; Bigelow 1992; BonJour 1992, 1998; Burge 1998; Butchvarov 1970; Katz 1998; Plantinga 1993). This effort has sometimes been coupled with an attack on naturalistic epistemology, especially in BonJour 1994 and Katz 1998. Such coupling is not surprising, because naturalistic epistemology is often associated with thoroughgoing empiricism and the rejection of the a priori. In this paper, however, I shall present a conception of naturalistic epistemology that is entirely compatible with a priori justification or knowledge. The resulting conception, I claim, gives us a better appreciation of the respective merits of the rational and the empirical, as well as a better understanding of how moderate epistemological naturalism comports, at least in principle, with moderate rationalism. This paper defends moderate rationalism; but it does not defend everything rationalists have often wanted, only what it is reasonable to grant them. The a priori is traditionally regarded as a type of knowledge, and sometimes as a type of truth. I shall follow the practice of recent discussants who treat the a priori primarily as a species of warrant or justification (I use these terms interchangeably). This has several advantages. First, it properly allows for the possibility that a belief might have a priori warrant but fail to be true, and hence fail to be a piece of knowledge. Second, it sidesteps, or at least marginalizes, the question of what else is required for knowledge beyond justified true belief. Third, it highlights the fact that unlike the necessary/contingent distinction, which is a distinction between types of truth, the a priori/a posteriori distinction is fundamentally concerned with sources of warrant or justification, not types of (true)

2 / Alvin I. Goldman

propositions. It is not wrong to use the term “a priori” as a predicate of propositions, e.g., as a predicate that applies to any proposition for which a person might have a priori warrant. But such a use is derivative from the central, epistemic sense of “a priori”. 2. What Is Naturalistic Epistemology? Many things can be meant by “naturalism” and “naturalistic epistemology”. Some forms of naturalism involve metaphysical theses—for example, the thesis that everything in the world either is physical or supervenes on the physical—and some forms of naturalism involve methodological doctrines— for example, the doctrine that proper methodology is purely empirical.1 I am concerned here with specifically epistemological forms of naturalism and shall therefore try, as far as possible, to skirt issues of metaphysical naturalism. Although rationalism might invite certain forms of metaphysical anti-naturalism, I shall remain largely neutral on these metaphysical issues. Epistemological naturalism itself comes in many varieties and flavors, however, none of which is uniquely correct or authoritative.2 I begin by first characterizing some radical forms of naturalistic epistemology which I do not endorse and then turn to a more attractive form of epistemological naturalism that I cheerfully embrace. The first form of radical epistemological naturalism will be called scientistic naturalism: (SN) Epistemology is a branch of science. The statements of epistemology are a subset of the statements of science, and the proper method of doing epistemology is the empirical method of science. This formulation of naturalistic epistemology is obviously drawn from Quine’s description of the subject in “Epistemology Naturalized” (Quine 1969). He there describes naturalistic epistemology as a “chapter of psychology and hence of natural science” (82). It studies a certain input-output relation involving a particular “natural phenomenon, viz., a physical human subject” (82). My principal dissatisfaction with this description of naturalistic epistemology is that no branch of empirical science, including psychology, takes on the (normative) tasks of specifying the criteria, conditions, or standards for justification and/or knowledge. But surely at least part of epistemology’s mission is to undertake these tasks. Thus, as many critics of Quine have pointed out (e.g., Kim 1988), his conception of naturalistic epistemology omits too much of what is distinctive of epistemology. Some might argue that the sciences of probability or statistics do address the question of standards of justification or warrant. However, these are not usually considered empirical sciences, and they are not the sciences Quine mentions in saying that epistemology should be a branch of science. Furthermore, it is controversial whether either of those sciences (or branches of mathematics) offers a theory of justified or warranted belief.

A Priori Warrant / 3

The second form of radical epistemological naturalism is what I shall call empiricist naturalism. I formulate it as follows: (EN) All justification arises from empirical methods. The task of epistemology is to articulate and defend these methods in further detail. Unlike (SN), (EN) has the virtue of addressing the nature of justification, which (SN) ostensibly dodges. (EN) also properly assigns to epistemology the task of clarifying appropriate epistemic methods. (EN) also differs from (SN) in not equating epistemology with some branch of empirical science. To that extent, it is a bit weaker than (SN). At the same time (EN) is stronger than (SN) insofar as it makes a more unequivocal commitment to empiricism. (SN) does not say that all justification arises empirically—in fact, it says nothing about justification at all. (SN) specifies that the study of epistemology must be empirical, but it is noncommittal on whether the warrant for all subjects is empirical. The obvious problem with (EN) is that it is far from clear that an adequate epistemology must hold that all warranted belief is empirical. Rejection of thoroughgoing empiricism, however, should not automatically exclude one from a defense of epistemological naturalism. There is a moderate conception of epistemological naturalism, I submit, that is perfectly compatible with non-empirical warrant.3 The naturalism I recommend is a fusion of two theses, the first concerning the generic source of epistemic warrant and the second concerning the nature of the epistemological enterprise. Here is my formulation of moderate naturalism: (MN) (A) All epistemic warrant or justification is a function of the psychological (perhaps computational) processes that produce or preserve belief. (B) The epistemological enterprise needs appropriate help from science, especially the science of the mind. Thesis (A) fits with the rather minimal metaphysical point that epistemic agents are natural phenomena, namely, physical organisms. It goes beyond this claim, moreover, in locating the source of warrant in the psychological or computational processes by which beliefs are formed and preserved. This thesis about warrant (originating, I believe, in Harman 1973 and Goldman 1979) corresponds to what Kitcher (1992) calls the psychologistic conception of epistemological naturalism. BonJour (1994) argues that psychologism is not a very distinctive ingredient of epistemological naturalism because “it is hard to believe anyone has ever disputed [it]” (290). “Minimal psychologism”, says BonJour, involves little or no departure from traditional epistemology. If “traditional epistemology” means the history of pre-20th-century epistemology, then I would agree. Indeed I have elsewhere emphasized that psychologism dominated the great sweep of historical epistemology (Goldman 1985, 1986: 6). But in our own era—that is, in the 20th

4 / Alvin I. Goldman

century—epistemological psychologism has been frequently attacked. During the positivist era epistemic concern with the “genesis” of a belief was considered a fallacy; Reichenbach (1938) and his successors drew a sharp distinction between the the context of discovery—to which causal questions were admitted as relevant—and the context of justification—to which causal questions were deemed irrelevant. Nor was the rejection of psychologism confined to the positivist period. Among recent epistemologists, Chisholm criticizes the causal approach (1989: 82–84) and erects standards of justification that ignore a belief ’s causes. Similarly, Lehrer labels it a “causal fallacy” to confuse a person’s reason for belief with the cause of his believing it (1990: 169). Thesis (B) of moderate naturalism merely says that the epistemological enterprise “needs help” from science. (MN) thereby differs from (SN) in not identifying epistemology with any branch or sub-branch of science, and in not limiting epistemology to narrowly scientific questions. But it shares with (SN) the idea that empirical science has important contributions to make to epistemology (more on these contributions later). The most salient feature of (MN) for present purposes is that it makes no commitment to any thoroughgoing form of empiricism. It leaves it entirely open that rational insight or rational apprehension might be among the sources of epistemic warrant. In particular, since rational insight or apprehension might be a variety of belief-generating causal process, the door is not closed to rationalistic warrant. Others who have defended a moderate conception of naturalism have not looked so kindly on the a priori. For example, Kitcher embeds the following thesis in his favored form of naturalism: “Virtually nothing is knowable a priori, and in particular, no epistemological principle is knowable a priori” (1992: 76). Similarly, Devitt identifies naturalism with the view that “there is only one way of knowing, the empirical way that is the basis of science (whatever way that may be). So I reject a priori knowledge” (1996: 2). In contrast to Kitcher and Devitt, the form of moderate naturalism I advocate—when combined with the conception of a priority I shall defend—does not take so dismissive an attitude toward the a priori. Certainly the formulation of epistemological naturalism contained in (MN) does not automatically exclude a priori sources of warrant as inevitably obscure, mysterious, occult, or epistemically disreputable. 3. Features of A Priori Warrant Certain properties have been historically associated with the a priori that might indeed be in tension with epistemological naturalism. Let us review these historically salient properties and ask whether they really must be retained in any sensible account of a priority. Following the lead of other recent writers, I argue that many of these properties—especially those posing the most serious conflict with naturalism—are inessential to a priori warrant and should be abandoned. Six properties are traditionally associated with a priori knowledge or warrant: (1) a non-experiential, i.e., non-perceptual, source or basis, (2) necessity,

A Priori Warrant / 5

(3) a subject-matter of abstract, eternal objects, (4) infallibility, (5) certainty, and (6) rational unrevisability (incorrigibility). The first of these properties—having a non-perceptual source—is unquestionably essential to a priority. The second of the six properties, necessity, is another firmly entrenched feature of a priority according to historical treatments. I am uncertain whether some sort of involvement in necessity is essential to the a priori. An appropriate restriction involving necessity is difficult to formulate, but I do not preclude the possibility of identifying one. It is a delicate matter that will be explored below but not firmly settled. The third property associated with the a priori is that its subject-matter should be abstract, eternal objects, such as numbers, universals, or meanings. Here I want to stick to my earlier resolve to stick to the epistemological questions concerning the a priori and avoid the metaphysical questions. Thus, I want to remain neutral on the issue of what the subject-matter of the a priori has to be. To be more precise, although I am willing to concede that only beliefs on certain topics or in certain domains will qualify as warranted a priori, I want to remain neutral on the question of what the truth-makers are in those domains. I want to be able to concede the possibility of a priori warrant about arithmetic without taking a position on what numbers are or must be. Given this desire for metaphysical neutrality, it is obviously unacceptable to make an abstract subject-matter a necessary condition for a priority.4 Concerning the three remaining properties, I join several contemporaries in rejecting the traditional notion that a priori warrant must possess them. I consider them in the order listed, beginning with infallibility. A mode of justification is infallible if its use always leads (perhaps necessarily) to truth. I interpret this to mean that whenever a person believes a proposition with sufficient a priori warrant, then that belief is true. By “sufficient” I mean enough warrant that it would qualify a belief for knowledge if it were true and satisfied other Gettier-averting conditions (see Casullo 1988). However, there are many historical and everyday cases that comprise counterexamples to infallibility—cases in which people had sufficient a priori warrant for beliefs that have subsequently been recognized as false. BonJour (1998: 111–112) presents three categories of such counterexamples. First, certain claims of mathematics and logic were at one time regarded as self-evident by all leading authorities but are now regarded as false, e.g., Euclidean geometry and naive set theory. Second, there are a priori claims of rationalist metaphysicians that cannot all be true because they conflict with one another. For example, reality cannot consist both of a system of timeless, windowless monads and also of one indivisible absolute mind. If these examples are not wholly convincing because it isn’t transparent that the foregoing beliefs were warranted, we may turn to BonJour’s third category. There are routine errors in calculation, proof, and reasoning. When adequate care is taken in such matters, a reasoner’s belief is presumably sufficiently justified on a priori grounds, but this still does not preclude all mistakes. The next traditional property of a priori justification is certainty: the highest possible level of warrant. According to the tradition, propositions known (or war-

6 / Alvin I. Goldman

ranted) a priori are self-evident, and self-evidence is typically taken to imply certainty. To require a priority to yield certainty is presumably to say that a priori justification comes in only one degree: the highest. So every a priori justified belief must be at least as well justified as any belief whatever; and if, as is commonly maintained, all empirical beliefs are less than certain, then all a priori beliefs must have greater justification than all empirical beliefs. Is this consequence plausible? Routine beliefs about mathematics and logic are presumably justified a priori. But if they are fallible, as conceded earlier, can they really be certain? Must they all be better justified than any perceptual belief whatsoever, including the belief that there is a telephone on the table before me? That is counterintuitive. The sixth traditional property of a priority is rational unrevisability. Quine famously wrote that “no statement is immune from revision” (1961: 43), and he associated revisability with empiricism. A priori statements, if there were any, would be statements that are rationally immune to revision. Putnam (1979) makes rational unrevisability a condition of a priority in saying that an a priori truth is one that “it would never subsequently be rational to reject no matter how the world turns out (epistemically) to be” (1979: 433). Similarly, Kitcher’s account of a priori knowledge implies that if a person has a priori warrant for proposition p at time t, then nothing he can undergo after t would rationally undermine this warrant: We can say that a proposition is unrevisable for a person at a time in case there is no possible continuation of that person’s experience after that time which would make it reasonable for her to change her attitude to the proposition. The explication makes it apparent why one might think that propositions which a person knows a priori are unrevisable for that person. If you have a priori knowledge that p, then you have an a priori warrant for a belief that p. Assuming that the warrant is available independently of time, then, given any continuation of your experience, you would have available to you a warrant which would continue to support belief. Hence, it would never be reasonable for you to abandon p in favor of its negation. Whatever trickery your experience may devise for you, you will always be able to undergo a process which will sustain the belief. (1981: 222)

Examining the unrevisability thesis, Casullo (1988) distinguishes strong and weak versions of unrevisability. The strong version says that if S is justified in believing that p a priori, then the statement that p is rationally unrevisable in light of any future evidence. The weak version says that if S is justified in believing that p a priori, then the statement that p is rationally unrevisable in light of any future experiential evidence. Casullo proceeds to give convincing counterexamples to each of these unrevisability constraints on the concept of a priori warrant.5 Charlie believes that p entails q after reflecting on a valid proof of this entailment, so presumably he has a priori warrant for this belief. But suppose there is a pseudoproof that p entails ;q such that if this pseudo-proof were brought to Charlie’s attention, he would not detect any flaws in it or be able to discount it in any other

A Priori Warrant / 7

fashion. Then his current a priori warrant for the original entailment is rationally revisable; this warrant could be defeated. Nonetheless, Charlie still possesses a priori warrant for the entailment now. This case refutes the strong unrevisability thesis. To refute the weak unrevisability thesis, Casullo presents an example involving defeating evidence of a neurophysiological sort. Like Charlie, Phil bases a logic belief on a certain proof he has carefully considered. But if Phil were to be apprised of a certain brain scan of his logical thinking, this information would undermine his warrant for trusting his own thought process—although we may suppose that the brain scan is flawed and his logical thinking is impeccable. Phil really does have a priori warrant for his belief, although that warrant is subject to empirical defeat. Notice that the mere fact that a source of warrant is subject to empirical defeat does not show that it itself is empirical. Introspective warrant for a belief about one’s bodily sensations might be defeasible by neurophysiological observations; but this does not imply that the original warrant is not introspective. For all of these reasons, properties (4), (5), and (6), which are often associated with a priority, should be peeled away from that concept. Once this surgery is accomplished, epistemological naturalism should be much less repelled by the prospect of the a priori. Infallibility, certainty, and unrevisability may indeed be unlikely bedfellows of naturalism, because the prospect is dim that any natural causal process is either infallible, certain, or closed to correction. But once these features are deleted from the defining characteristics of the a priori, naturalists can find it much more palatable. Another feature of rational insight, however, might frighten off naturalists. This is the perceptual model of rational insight, in which the objects of rational insight are somehow cognized in a fashion analogous to the perception of physical objects. Perception is a causal process, in favorable cases, a process that causally connects a perceived object with the perceiver’s mental experience. If rational insight is understood on this model, it must consist in a causal connection between the realm of rationally knowable objects and the knower’s cognitive awareness. But it is highly doubtful, from a naturalistic perspective, that such causal connections could obtain. Benacerraf (1973) crystallized this problem in the domain of numbers. If numbers are Platonistic entities, can they really have a causal connection with people’s mental lives? This problem seems particularly threatening to the form of naturalism adopted here, because (MN) endorses a causal theory of warrant. How can this form of naturalism be reconciled with a priori knowledge? A crucial step in the reconciliation is to distinguish two types of causal processes, what I shall call intra-mental processes and trans-mental processes. Intramental processes occur wholly within the mind; trans-mental processes include links that are external to the mind as well as links that are internal. Warrantconferring processes, as envisaged by (MN), are intra-mental processes; they don’t encompass objects outside the mind (although the contents of their constituent states may refer to such objects). Thus, a priori warrant does not require the sort of trans-mental, perception-like process that Benacerraf was discussing.6

8 / Alvin I. Goldman

If a priori warrant does not require a perception-like causal connection between mental apprehensions and extra-mental realities that these apprehensions are about, how can they (non-accidentally) get the truth about the extra-mental realities (which is required for knowledge)? Bigelow (1992) suggests one possibility: necessities in the world might be “reflected” in the minds of the agents who are seeking to understand that world. “[A] necessary link between representations may . . . mirror a necessary link in the world, between the things which they represent. This harmony between representations and things represented is the source of the a priori character of mathematical knowledge.” (1992: 155) Here again we see that a priori warrant and a priori knowledge are not committed to assumptions that will inevitably repel a conscientious naturalist. Of course, if the truth-makers of mathematical propositions are abstracta, that will presumably worry a metaphysical naturalist. But as long as we focus on epistemological naturalism, which sets metaphysical issues aside, there is no cause for concern. Let us review where we stand. Properties (4), (5), and (6) have been carefully peeled away from the a priori, and property (3) has been scrupulously set aside. Does this leave us with a precise enough specification of a priori warrant, precise enough to distinguish it from other forms of warrant? We are left with two ingredients for distinguishing a priori from other forms of warrant, and it is unclear whether these two ingredients are adequate to the task. The first ingredient for characterizing a priori warrant is a negative feature: the absence of an experiential or perceptual basis of belief. Even this feature, of course, needs to be clarified if it is to serve as a necessary condition for a priority. As is often pointed out, the ability to comprehend a proposition may require perceptual experience to learn its constituent concepts. We don’t want such perceptual concept-learning to count against the possibility of a priori warrant for belief. This much understood, we must next ask whether a non-experiential, nonperceptual basis for belief is a sufficient condition of a priority. Burge contends that it is: A warrant . . . is apriori if neither sense experiences nor sense-perceptual beliefs are referred to or relied upon to contribute to the justificational force particular to that warrant. (1998: 3)

This is a negative characterization of the a priori: a warrant is a priori if it is not perceptual. Such a negative condition, however, will not do. To take just two examples, memory warrant and introspective warrant are both forms of nonperceptual warrant, but neither is a priori. A belief formed purely by memory, as when one recalls having had a certain (non-sensory) thought, does not have perceptual warrant; but neither is its warrant of the a priori variety. Similarly, introspection can give rise to warrant, but its type of warrant is neither perceptual nor a priori. Introspection should not be regarded as a species of perception, especially for present purposes, because it has no distinctive type of sensory experience associated with it. Of course, many objects of introspection—e.g., pains,

A Priori Warrant / 9

itches, and tickles—have sensory qualities, but introspection per se does not. One can introspect thoughts without any accompanying sensory quality. So one cannot equate a priori warrant with non-perceptual warrant.7 Instead of marking a priori warrant by the absence of sensory phenomenology, perhaps it can be marked by a positive form of phenomenology, a phenomenological mode unique to the a priori. This move was popular among traditional rationalists, and still finds favor among a few contemporary theorists. According to Plantinga (1993), when one sees that a proposition is necessarily true, one forms a belief “with that peculiar sort of phenomenology with which we are all acquainted, but which I can’t describe in any way other than as the phenomenology that goes with seeing that such a proposition is true” (106). Is this approach well-motivated? Is there a single distinctive phenomenology that accompanies all (purely) logical thought, mathematical thought, and other forms of cogitation that rationalists like to subsume under the a priori? This seems very doubtful. I do not reject the idea that rational thinking, when fully conscious, has phenomenological dimensions (see Goldman 1993). But I see no reason to suppose that a single distinctive form of phenomenology accompanies every a priori warranting process. Reliance on phenomenology, then, does not seem to capture the a priori. Can the situation be improved by adding the necessity ingredient? The simplest appeal to the necessity ingredient would be the following condition: A belief has a priori warrant only if the proposition believed is necessary. One problem with this proposal is that Kripke (1980) has argued that a person can have a priori warrant even for contingent propositions, e.g., “the standard meter stick is one meter long”. This example does not convince everyone, however, so I shall not press it. A second difficulty is more decisive. Since a priori warrant is not infallible, it is possible to have a priori warrant for propositions that are not in fact true and hence not necessary. A different way of bringing necessity into the picture is through the following condition: A belief in p has a priori warrant only if the doxastic agent also believes that p is necessary. Is this condition acceptable? It seems to me implausible. Ordinary people often believe arithmetic propositions without believing them to be necessary. Unlike philosophers, they do not reflect on matters of modal status. The proposition 2 1 5 5 7 just strikes them as true, not necessarily true. So there are many cases in which this condition is not met. A third possible way to introduce necessity into the picture is more promising: A belief in p has a priori warrant only if p belongs to a family or domain of propositions each of which is either necessarily true or necessarily false.

10 / Alvin I. Goldman

Propositions of arithmetic comprise a family of propositions that meet this condition, and so do propositions of logic. They are precisely the sorts of propositions for which people can have a priori warrant. Even if we accept this necessary condition of a priority, however, it does not combine with the non-perceptual-basis condition to yield a sufficient condition of a priori warrant. That is to say, the following proposal is not correct: A belief in p has a priori warrant if (1) the basis for this belief is nonperceptual and (2) p belongs to a family or domain of propositions each of which is either necessarily true or necessarily false. Obviously, a belief can meet the first of these conditions without being warranted a priori, indeed, without being warranted at all. For example, the belief might be based on mere wishful thinking, or sheer guesswork. Even beliefs in truths of logic and arithmetic might have this kind of source. The combination of nonperceptual basis and appropriate modal status does not suffice to confer a priori warrant on such beliefs. We have not succeeded, then, in explicating the concept of a priori warrant. Does that scuttle epistemic rationalism? No. The fundamental idea of a priori warrant is the idea of purely rational warrant, warrant based on “pure thought”. This intuitive idea has not been shown to be devoid of merit. We should not jettison the assumption that it refers to a definite phenomenon (or set of phenomena) simply because we cannot yet provide a fully illuminating characterization of it. For one thing, it may be impossible to elucidate what a priori warrant is until we have an account of warrant in general. So let us turn next to that task. 4. Reliabilism, Warrant, and the A Priori Thesis (A) of moderate naturalism identifies psychological processes as the generic source of all epistemic warrant. This is hardly a complete theory of warrant, however, so more must be said about warrant before we can fully assess the prospects for a priori warrant from the perspective of moderate naturalism. Here I shall defend a theory I have put forward earlier, a version of process reliabilism. Process reliabilism fits very comfortably, of course, into the category of naturalistic approaches specified by (MN)(A). I have not built it explicitly into (MN)(A) because many other psychological-process accounts of warrant are in principle available. Any such theory is plausibly regarded as naturalistic—at least as having one naturalistic component. Reliabilism does not exhaust naturalistic approaches to warrant, though it is the only such approach I view as correct. The version of reliabilism I wish to endorse here is two-stage reliabilism (for details, see Goldman 1992). Two-stage reliabilism offers the following reconstruction of our standards for epistemic warrant. During various periods of a community’s evolution, the criterion of reliability is applied to various beliefforming processes and methods, which are individuated in some rough-and-

A Priori Warrant / 11

ready fashion.8 Certain belief-forming processes pass the test of reliability—or are thought by the community to pass it—and other belief-forming processes do not pass it. Basing expectations about future battles on the features of animals’ entrails, for example, was regarded as a sound method in some communities. When a process or method is judged to have a high proportion of true outputs, it is viewed as a warrant-conferring process or method. Thus, the root or “Ur” criterion of warrant is reliability of belief-formation. (Notice that the criterion appealed to is reliability, not judged-reliability or believed-reliability. Actual reliability is the criterion that the community tries to apply, though it may fail to apply that criterion correctly.) I shall call this first stage of the epistemological story the standard-selection stage, because it involves the selection of approved epistemic standards, viz., the approved belief-forming processes or methods that confer epistemic warrant. The second stage of the epistemological story is the standard-deployment stage. In this stage, members of the community apply the chosen standards by judging whether individual beliefs (either actual or hypothetical) are warranted as a function of whether they are arrived at (or sustained) by approved processes or methods. According to my hypothesis, although standards are chosen by the ( judged) reliability of processes in the actual world, the chosen standards are then applied “rigidly”. That is, each of the approved processes or methods is deemed warrant-conferring in any possible world in which it operates. (This will only be relevant when the judged belief is a hypothetical belief, e.g., one supplied in a philosophical example.) This reconstruction helps account for intuitive judgments about justifiedness that pose problems for other versions of reliabilism, e.g., the justifiedness of beliefs in Cartesian demon worlds (see Goldman 19929 ). The following sorts of processes are examples of approved belief-generating processes, at least in our epistemological community: perceptual processes in the several sense modalities, remembering, introspecting, and (many forms of ) reasoning or calculating. I have tried to reconstruct the way in which communities and individuals select and deploy their standards of warrant or justification. But, it will be asked, when are beliefs really justified, as opposed to being held justified by this or that community? A natural response is: a belief is “really” justified if and only if it results from processes (or methods) that really are reliable, and not merely judged reliable by our present epistemic community. Thus, we should be prepared to change our standards of justification if we find reason to believe that processes or methods previously thought reliable are really unreliable, or if we find that previous ways of individuating or grouping modes of belief-formation do not cut the cognitive mind “at its joints,” which could affect assessments of reliability. Now one family of these justification-conferring processes is a good candidate for conferring a priori warrant, i.e., the family of reasoning or calculational processes. If this proposal is accepted, then beliefs formed exclusively by such processes—for example, beliefs formed by purely arithmetic or logical reasoning—would be warranted a priori. Furthermore, we might expand the ambit of the a priori by following the lead of BonJour. BonJour suggests that

12 / Alvin I. Goldman

instead of restricting a priori warrant to beliefs, we might also allow inferences to have a priori warrant (BonJour 1998: 4– 6). Adapting this idea to my process approach, we may say that certain sorts of processes, including inferential processes, are a priori warranters. (Non-inferential processes might also qualify as a priori warranters, e.g., processes for apprehending “basic” necessary truths, which are not derived inferentially from other such truths.). This proposal is not meant to imply that whenever an agent uses a process that is an priori warranter, any belief-output of the process is wholly a priori. On the contrary, if one starts with a set of believed premises that originate in perception and then applies an inferential a priori warranter to that set of beliefs, the resulting conclusion belief does not have pure a priori warrant. Nonetheless it seems instructive to say that such a conclusion belief has an element or component of a priori warrant, simply because there is one strand of its warrant that is a priori. It should be noted that when dealing with inferential processes, we need to qualify the reliabilist criterion slightly (see Goldman 1979). Here the criterion of warrant conferral should not be simple reliability but conditional reliability. An inferential process should not be expected to have a high truth-ratio of output beliefs when the input beliefs are false. It is only necessary that its output beliefs have a sufficiently high truth-ratio for those cases in which the input beliefs (premises) are all true. This conditional understanding of reliability for inferential cases will be presumed in the remainder of the discussion. There are several objections I can anticipate to my entire approach. First, epistemological internalists will undoubtedly want to object to the fundamental reliabilist contours of the approach. Setting aside the distinctive features of twostage reliabilism, internalists characteristically deny that de facto reliability of belief-forming processes suffices to give them warrant-conferring power. A belief does not attain justificational status merely by resulting from reliable processes; that makes the believer’s justification too “external” to him. Instead, the believer must also possess a metajustification: he must be justified in believing that his processes are reliable. A leading proponent of this metajustification requirement has been BonJour (1985). Interestingly, when it comes to a priori warrant, BonJour explicitly rejects the appropriateness of metajustification (1998: 142–147). In defense of this move, he says that a metajustification would have to be either a priori or empirical. If it were a priori, it would be circular; if it were empirical, the a priori would lose its presumed status as an “autonomous” mode of justification. I shall not explore in detail these reasons for rejecting the metajustification requirement, though I find them unconvincing. But I do wish to register surprise that BonJour does not recognize this position as a reversal of his long-standing dissatisfaction with externalism. It is curious that he continues to think of himself as a champion of internalism (see 1998: 1, 96–97) even with respect to the a priori, despite his abandonment of the metajustification requirement. My own reason for rejecting the metajustification requirement was given in Goldman 1979.10 It is too demanding a constraint on justification that an ordinary

A Priori Warrant / 13

cognizer should possess such a metajustification. This is particularly so in the domain of the a priori. It is very unlikely that someone who has never studied philosophy could produce a satisfactory justification for the reliability of his inductive or deductive inference procedures. To conclude from this, however, that ordinary, philosophically untrained people have no inferential warrant would be a dramatic capitulation to skepticism. A different kind of objection to the reliabilist approach to a priori warrant would be an objection to its contingent character. Is it sufficient for an inferential process to have a priori warrant that it be (conditionally) reliable in the actual world? Isn’t a stronger requirement appropriate, namely, that it be necessarily reliable: reliable in every possible world?11 Such a requirement would not constitute a retraction of our earlier insistence that a priori warrant be compatible with fallibility. The current proposal does not require perfect reliability in every possible world; it merely requires a “sufficiently” high truth-ratio in every possible world. This proposal feels quite appropriate for mathematical and deductive reasoning processes, but it looks excessive for inductive processes. Yet it is definitely plausible to hold that inductive reasoning processes (of certain types, at any rate) can be a priori warranters. Clearly, inductive connections between propositions are not epistemically accessible by perception, memory, or introspection; they seem to be accessible only by some species of “reason”. The problem is that for any inductive process mapping premise beliefs into conclusion beliefs, there will be possible worlds in which the process is unreliable (that is, has very low reliability). However, it is possible to construe inductive inference as tracking probabilistic relations between premises and conclusions, and these probabilitistic relations might hold in all possible worlds.12 So forms of inductive reasoning might be candidates for a priori warranters even if necessary reliability is required for a priori warrant. For this reason, I am not ready to dismiss the demand for necessary reliability as a condition on a priori warranters. An alternative approach, however, would say that since we have already admitted the fallibility of a priori warranting processes, error possibilities are already on the table in the sphere of the a priori. If a priori warrant is compatible with error in the actual world, why shouldn’t it be compatible with further error—perhaps even unreliability—in other possible worlds? So I shall not try to resolve this matter fully. Still another possible worry about our account is whether it can be suitably supplemented to handle a priori knowledge. Granting that the primary notion of a priority is that of warrant rather than knowledge, an account of a priori warrant must be “upgradable” into an account of a priori knowledge. Does the reliability theory have such a potential?13 The simplest way to complete a reliabilist account of knowledge is to add both the standard truth requirement and a no-relevantalternatives requirement.14 Admittedly, the no-relevant-alternatives condition will not do much work whenever the beliefs in question are beliefs in necessities, because counterfactuals involving the negations of necessities are either trivially true or at least difficult to evaluate. So it will be hard to show that anyone ever violates the no-relevant-alternatives condition if he already believes a necessary

14 / Alvin I. Goldman

truth. For the very same reason, however, it does not seem that anyone with a true, a priori warranted belief in a necessity genuinely faces any further threat to knowledge. So there simply is no real “work” for the no-relevant-alternatives condition to do in that class of cases. In the final segment of this section, I want to raise a further complication for the account of a priori warrant. I earlier introduced, with little comment, a distinction between belief-forming processes and methods. Now I want to say more about that distinction and its relevance to the theory of a priori warrant. By a “process” I mean something that is part of a person’s fundamental cognitive architecture. By a “method” I mean something that is not part of one’s fundamental cognitive architecture, but something learned, typically by cultural transmission. For example, a truth-table procedure for determining which sentences are tautologies is unlikely to be a process in the foregoing sense; it is unlikely to be part of a person’s fundamental cognitive architecture. Someone might acquire such a method by learning, however, and this method would presumably be a paradigm case of a necessarily reliable method. Is such a method an a priori warranter? I think not. Consider the case of Carroll’s tortoise (Carroll 1895). He apparently has no native, intuitive power to detect logical relationships. If he were persuaded to accept such relationships on Achilles’s authority, his belief in such relationships might be warranted; but it would not qualify as a priori warrant. Let us return to human beings and to the truth-table method. By my lights, two cases must be distinguished. In one case Harry learns the truth-table method from Ellen, who simply explains how to use it without explaining why it is (necessarily) reliable. Harry simply accepts its reliability from Ellen on trust; he does not use his prior reasoning powers to “see” that it is reliable. In a second case Harry learns the truth-table method from Eileen, who explains why the method is (necessarily) reliable, an explanation that Harry fully comprehends and appreciates in virtue of his prior reasoning powers. In the first case it seems clear that the truthtable method of forming beliefs about tautologies is not an a priori warranter. For one thing, the method is acquired in part by perception (of Ellen’s testimony), and that perception is not an incidental or eliminable feature of Harry’s acceptance of the method.15 In the second case Harry seems to have a priori warrant for his belief that the method is (necessarily) reliable, because he himself determines its reliability by pure reasoning powers. This might suffice to make the method itself an a priori warranter. But I am not fully convinced of this. Guided by historical treatments of the a priori, I am inclined to say that only basic mental faculties have the power to confer a priori warrant. In other words, only “processes” in the present sense, are a priori warranters. However, I might be prepared to concede that methods can also become a priori warranters if their (necessary) reliability is determined by a priori means. In either case, it turns out that the underlying source of all a priori warrant must reside in appropriate processes, which are features of our fundamental cognitive architecture. Do people actually have such processes? If so, exactly which such processes do they have?

A Priori Warrant / 15

5. Naturalism, Cognitive Science, and the A Priori The form of epistemological naturalism I have endorsed, moderate naturalism, claims that epistemology needs help from science, especially the science of the mind; but I have not yet fully explained why science should enter the picture. The remarks at the end of the previous section indicate the direction of the rationale, as applied specifically to the question of the a priori. But let me back up a little to give the rationale more fully. Two-stage reliabilism is a reconstruction of ordinary people’s conception and standards of justification. As indicated, however, epistemology should not rest content with lay people’s standards of warrant. There are several reasons why a scientifically-infused epistemology might plausibly part company with folk epistemology. First, the folk presuppose some categories of belief formation, some types of belief “sources”. But it is doubtful that the folk have landed upon an optimal classification of belief-forming processes. There is ample reason to suspect that scientific psychology is needed to cut the operations of cognition “at their joints”. This is one reason to seek help from psychology. Second, according to two-stage reliabilism, the folk have somehow identified certain belief-forming processes as having relatively high truth-ratios and other belief-forming processes as having relatively low truth-ratios. But it is debatable whether all of these assessments are accurate. A scientific understanding of exactly how the indicated processes work might lead us to a different assessment of their truth-producing capacities ( just as the study of perceptual illusions can shed light on matters of reliability). Third, if philosophers want to distinguish a priori from other types of warranters and specify which are which, then, for reasons sketched at the end of the last section, psychology needs to tell us which processes are parts of our fundamental cognitive architecture and which are used in acquiring new reasoning methods. Finally, rationalist philosophers have identified a single kind of putative knowledge source that they call (roughly) “rational intellection”. A more scientific approach to cognition might make fruitful headway into this terrain, not by rejecting such a source altogether, but by revealing that the supposedly single source of rational insight is really comprised of an assemblage of different mechanisms. I begin with the subject of numerical or arithmetic cognition. Might there be psychological evidence that supports the existence of a psychological capacity for reliable (even necessarily reliable?) numerical cognition? Yes, there might; indeed, there actually is such evidence. I should preface my exposition of this evidence by pointing out that the possible capacity I shall discuss does not speak to the apprehension of numbers in the full-blown Platonistic sense. It may only be a capacity to discern relations of numerosity among sets of objects. From my point of view, nothing more is needed for a priori numerical cognition. In the last 10–20 years it has been shown that both animals and human infants as young as five months are sensitive to number (see Wynn 1992a). Rats are able to determine the number of times they have pressed on a lever, up to at least 24 presses. Birds have similar abilities. Canaries were trained to select an object

16 / Alvin I. Goldman

based on its ordinal position in an array. Out of 10 cubicles spaced along a runway, they had to walk along the runway and choose the cubicle that held, say, the fifth aspirin (Pastore 1961). In another study, Church and Meck (1984) showed that rats can compute small numerosities, such as two plus two. Similar findings have now been made among human infants. Using the standard technique of gauging surprise by length of looking time, Wynn (1992b) found that five-monthold infants can correctly detect elementary arithmetic relationships, such as 1 1 1 5 2 and 2 2 1 5 1. Infants saw arithmetic operations of placing or removing (adding or subtracting) items from a display area, though a screen initially prevented them from seeing the result of the operations. When an “impossible” result was revealed to them (e.g., 2 2 1 5 2), the infants manifested surprise (as measured by comparative looking time). All this evidence points to innate arithmetic powers among animals and human infants. How might they execute their arithmetic computations? One hypothesis, directed primarily at animals but possibly involving children as well, is that a single mechanism underlies the ability to determine numerosity and also the ability to measure duration. This is the so-called “accumulator theory” advanced by Meck and Church (1983), based on a model by Gibbon (1981). The proposed mechanism works as follows: [A] pacemaker puts out pulses at a constant rate, which can be passed into an accumulator by the closing of a mode switch. In its counting mode, every time an entity is experienced that is to be counted, the mode switch closes for a fixed interval, passing energy into the accumulator. Thus, the accumulator fills up in equal increments, one for each entity counted. In its timing mode, the switch remains closed for the duration of the temporal interval, passing energy into the accumulator continuously at a constant rate. The mechanism contains several accumulators and switches, so that the animal can count different sets of events and measure several durations simultaneously. The final value in the accumulator can be passed into working memory, and there compared with previously stored accumulator values. In this way the animal can evaluate whether a number of events or the duration of an interval is more, less, or the same as a previously stored number or duration that is associated with some outcome, such as a reward. (Wynn 1992a: 323–324)

Evidence that the same mechanism underlies timing as well as counting processes includes the fact that methamphetamine increases rats’ measure of duration and of numerosity by the same factor. Another nativist theory of numerical competence has been advanced by Gelman and colleagues (e.g., Gelman and Gallistel 1978, Gelman and Greeno 1989). She has proposed that young children possess an innate concept of numbers governed by three counting principles that define correct counting. One principle states that items to be counted must be put into one-to-one correspondence with a set of (innate) mental counting tags; a second principle says that these tags must be used in a fixed order; and the third principle states that the last number tag used in a count represents the cardinality of the counted items.

A Priori Warrant / 17

Wynn points out that in the accumulator model, numerosity is inherently embodied in the structure of the hypothesized representations. The relationships between the representations exactly reproduce the relationships between the quantities they represent. For example, four is one more than three, and the representation for four (the magnitude of fullness of the accumulator) is one more increment than the representation for three. Thus, a mechanism of this sort would mirror the structure of the facts that are represented. This makes it a kind of mechanism well suited for reliable numerical calculation. As Wynn also points out, addition could be achieved in an accumulator mechanism by transferring the energy from two accumulators into an empty third accumulator. These are only two theories, of course, of numerical cognition in infants and animals. I am not claiming that either theory has been proved to be correct. But if either theory were correct, there would be an innate mechanism of a priori numerical cognition that at least approaches the desiderata of some rationalist philosophers. There is nothing to suggest that such mechanisms would have any sort of trans-cognitive causal relation with numbers construed as abstract entities. And there might not be any distinctive phenomenology accompanying the operation of either mechanism (although, on the other hand, there might be such a phenomenology). But both of these traditional desiderata were already dismissed here as inessential to a priori cognition. I turn next to deductive logic, a subject to which psychologists have devoted considerable attention, yielding a proliferation of theories. Certain of these theories comport fairly nicely, at least at first blush, with the claims of a priorist philosophers. One going psychological theory is that ordinary people have something like natural-deduction systems built into their heads (Rips 1994, 1995; Braine, Reiser, and Rumain 1984), quite possibly innately (Macnamara 1986). They possess at least some sound natural-deduction rules and have operating systems that enable them to apply such rules to proffered sets of sentences to assess their derivability. When a person “intuits” that P is derivable from P-AND-Q, she may be deploying a sound mental rule of inference or proof. Since the deployment of such a mental rule would constitute a reliable process—even a necessarily reliable process—it would pass the test of a reliabilist conception of the a priori. Another approach to mental logic claims that people execute logic tasks by constructing mental models and seeing whether a proffered conclusion is true in all models in which the premises are true (Johnson-Laird and Byrne 1991). This approach might also sit comfortably with a priorist conceptions, although the process of constructing appropriate models is in general quite difficult. Only for certain types of logic tasks would the process be successful (the same holds for the proof-theoretic approach). Neither of these approaches, however, has clear-cut support from a majority of current researchers. Other psychological approaches to logic cognition offer frameworks quite alien to traditional epistemological thinking and alien to the prospects of a priori warrant in this domain. One approach suggests that skills at conditional logic, for example, are not at all innate, but instead are learned by an

18 / Alvin I. Goldman

inductive process from particular contingencies between events (Holland et al. 1986, 281–282). Support for this viewpoint is adduced from the fact that in a famous type of test for conditional reasoning, the Wason selection task, there are well-known “content effects”. Subjects presented with a form of the task involving arbitrary and unfamiliar subject-matter perform quite poorly, whereas subjects given the same task (same from a logical perspective) employing familiar subject-matter perform quite a bit better. A possible explanation for these content effects is that people lack domain-general competence at conditional logic, but acquire a certain level of competence through an inductive learning process.16 Adherents of this approach might concede that a few people—those adequately trained in pure logic—ultimately achieve wide-ranging competence. But this only shows that people can acquire reliable methods (in my terminology) for executing formal logic tasks. It does not show that anyone has innate processes of the sort that may be needed for a priori warrant in matters of deduction. I have, of course, left the door open to the idea that learned methods might be a priori warranters. But they can serve as a priori warranters, I suggested, only if the learning processes themselves can demonstrate that deductive methods are necessarily reliable. It is not clear how this could be achievable under the inductive learning approach currently under discussion. A recently developed approach to logic cognition, and perhaps the most extreme approach, endorses the idea that there are innate or genetically endowed reasoning mechanisms. But this approach, championed by Cosmides and Tooby (Cosmides 1989, Cosmides and Tooby 1994), denies that our genetic endowment includes any sort of abstract or domain-general capacity for deductive reasoning (or for inductive reasoning either). On the contrary, it hypothesizes that our genetic cognitive endowment consists of highly specialized “modules”, each of which is dedicated to a narrow cognitive task that proved adaptive in our evolutionary history. They specifically argue that the capacity to reason correctly with conditionals is restricted to tasks that involve the detection of “cheaters”, that is, the detection of people who have violated the terms of a social contract or social exchange. According to Cosmides and Tooby, lay people only succeed at those permutations of the Wason selection task in which the logically correct response coincides with effective cheater-detecting choices. For example, subjects are given problems in which they have to decide which information-revealing cards need to be examined in order to determine whether a conditional sentence holds true in a certain situation. According to Cosmides, people succeed at this task only when the content of the conditional is a social-contract type of rule on which somebody might “cheat”, i.e., when the rule’s content is of the kind: “If you take the benefit, then you pay the cost”. This thesis may sound far-fetched to philosophical ears, but it has at least initial support from intriguing experimental evidence. For example, Cosmides (1989) found that when subjects were given switched social contract rules, i.e., rules of the form “If you pay the cost, then you take the benefit”, they chose cards appropriate to cheater detection rather than to the principles of logic. Logic re-

A Priori Warrant / 19

quired the choice of a certain pair of cards (in a group of four) to determine whether or not the specified conditional holds. Instead, given the switched socialcontract rules, subjects ignored the logically correct cards and chose those suited to cheater detection.17 If Cosmides and Tooby were right about logic cognition in general, the prospects for a priori warrant in the deductive reasoning domain would indeed be dim. True, people would have innate processes enabling them to form logically accurate beliefs for some tasks. But these processes would also lead to logical errors in other tasks; so it is questionable whether the processes would be sufficiently reliable to qualify as epistemic warranters. It is even more doubtful that they would meet the condition of necessary reliability, which I have (tentatively) endorsed for a priori warrant. All this would apply if the Cosmides-Tooby story were correct, but I am pretty dubious about it, at least as a full story of human logical competence. Cosmides and Tooby have not, to my knowledge, addressed human abilities to solve simpler logic tasks, e.g., tasks involving AND-elimination, AND-introduction, or IF-elimination (Modus Ponens). Rips (1995) provides evidence for substantial lay competence with these sorts of reasoning principles. So I am not arguing that empirical studies in fact imperil the prospects for a priori warrant in the domain of deductive logic. I am only arguing that empirical studies could in principle imperil such prospects. A final domain that might be investigated here is that of inductive or probabilistic inference. A good deal of psychological research has raised questions about the capacity of naive judgment to conform with principles of probability. Doubt on this score has been the main message of the influential “biases and heuristics” approach of Tversky and Kahneman (1974, 1983). However, a careful treatment of this literature and some of the critical discussion of it (e.g., Gigerenzer 1991) poses difficult issues, both about the interpretation of probability and about the interpretation of a large body of psychological research. There is also a special problem of assessing the import of these findings in terms of the reliabilist framework advanced here. So although this terrain might prove fertile ground for illustrating my earlier points about the pertinence of empirical psychology, I shall rest content with the earlier illustrations. There is one additional point, however, that is worth emphasizing. The traditional rationalist picture, I think, is that a priori insight or apprehension is one homogeneous type of cognition, ostensibly intended to explain knowledge of logic, mathematics, meaning relationships, and so forth. But if the accumulator device, say, is a good example of the sorts of mechanisms and processes that lie at the roots of rational intellection, there is unlikely to be a single faculty of rational intellection. Obviously, the accumulator device is only useful for counting and related numerical operations. It is wholly incapable of subserving any of the other functions traditionally ascribed to rational insight. So there would have to be other mechanisms and processes that subserve those functions. Thus, a sober cognitive science of rationality might have to postulate not only a numerical device, but separate devices for deductive logic, inductive or probabilistic

20 / Alvin I. Goldman

relationships, and so forth. There would be processes of intellection1 , intellection2 , intellection3 , and so forth. In this fashion, naturalistic epistemology—that is, epistemology that exploits the resources of cognitive science—would in some measure support the ideas of traditional rationalism, but would also transform rationalism into a form entirely respectable by contemporary scientific lights. 6. Philosophical Analysis, Intuitions, and Empirical Science In this section I consider some objections to my thesis of moderate naturalism. I have defended moderate naturalism from the vantage-point of process reliabilism, but is this an adequate defense? Some critics might complain that they don’t accept process reliabilism as an adequate analysis (or reconstruction) of justified belief. So why should they be persuaded by my brief for the role of science in epistemology? Other critics might complain that although process reliabilism is a credible theory, the basic epistemological task of establishing process reliabilism is itself a task for conceptual analysis; and I have done nothing, they would add, to show that the enterprise of conceptual analysis needs help from empirical science. Furthermore, they would continue, conceptual analysis is really a purely a priori enterprise, because it rests fundamentally on a priori intuition. Conceptual analysis consists of offering accounts of certain concepts, in the case of epistemology, epistemological concepts. These candidate accounts are tested by seeing whether their implications accord or fail to accord with our intuitions. Thus, a priori intuition is the method for assessing philosophical analyses. No empirical science need be consulted. Now I agree that intuitions about cases play precisely the evidential role in testing conceptual analyses that the foregoing argument maintains. But I am not convinced that this establishes that philosophical methodology is, or should be, wholly a priori. (The reply I proceed to offer is adapted from a more extended treatment of conceptual analysis presented in Goldman and Pust 1998.) I regard intuitions—i.e., intuitional states of the type in question—as conscious, spontaneous judgments. They either are beliefs or they readily give rise to beliefs when not overridden by independent information. I will understand them in the latter fashion, as states that tend to generate beliefs. As used by philosophers, intuitions are typically about whether particular cases are or are not instances of a particular concept. For example, someone reflecting on one of Gettier’s (1963) cases may have the intuition that Smith’s true justified belief that p is not an instance of knowledge. One important question is how we should construe this talk of “concepts”. The construal I advocate, for present purposes, is that a concept is some sort of non-conscious psychological structure or state that is distinctively associated with the cognizer’s deployment of a certain natural-language predicate, in this case, the predicate “knows”.18 Such structures or states have contents, and hence particular examples—either actual or hypothetical examples— either instantiate those contents or fail to instantiate them. Now when someone has an intuition that case C is (or is not) an instance of concept K (as he represents

A Priori Warrant / 21

or understands K), does that intuition qualify as evidence for the claim that C is an instance of K? In other words, does the occurrence of this intuitional state justify the belief to which it naturally gives rise, that C is an instance of K? Certainly this is what philosophers generally assume and I have no quarrel with it. Furthermore, it meshes perfectly with my process reliabilist approach to justification. The reason such a belief is justified is because the process by which it is generated is a generally reliable process. Usually, when beliefs are formed via intuitions in this sort of way, those beliefs are true: C does instantiate the content of the person’s concept K. Admittedly, intuitions can sometimes go astray. When an intuition is produced by the person’s semantical theory about concept K, rather than as an expression of K itself, then the intuition can easily be mistaken. But if the intuition arises from a spontaneous application of concept K itself, the intuition will usually be right, as will the belief it generates. The fact that this is a reliable process accords quite well with the standard philosophical assumption that such intuition-based beliefs (about particular cases) are warranted beliefs. So far I take myself to be in substantial agreement with the imagined critic. But the critic proceeds to assert that the intuition-based beliefs in question are not only warranted but are warranted a priori. This is what he means in saying that the evidence for conceptual analyses resides in a priori intuition. This is where I part company with the critic. I do not see why this warrant should be considered an a priori form of warrant. On the view I would propose, intuitive “access” to certain features of a (non-conscious) concept—for example, access to the fact that it is instantiated by a certain case C—is akin to perception. Better yet, it is a form of interoception. Since a concept, on my construal, is a non-introspectible structure or state of one’s cognitive system, it is part of oneself. An intuition that reflects or reports something about that structure or state is analogous to a sensation that reflects or reports upon some condition of the body, e.g., the position of a limb. When intuitions are understood in this fashion, as analogous to inner perceptual experience, there is no longer any reason to think of them as a priori warranters. They are indeed warranters, just not a priori warranters.19 Another reason to reject the suggestion that intuitionally generated beliefs are warranted a priori is that the propositional contents of intuitions are neither necessarily true nor necessarily false. If I am right, the contents of these beliefs are of the form, “C satisfies the content of my concept that I express with the predicate ‘K’ ”. That sort of proposition is neither necessarily true nor necessarily false, since it is quite contingent what content one does associate with ‘K’. So if a priori warranters must be processes directed at domains involving necessity, we have an additional reason to reject the idea that philosophical intuitions—i.e., intuitions about concept instantiations—have a priori warrant. Of course, this is not to say that they have empirical warrant either. I am not claiming that the intuitional warranting of propositions about concept satisfaction is a form of external perception. Much of what I have said about concepts and their relationship to intuitions is similar to the views of Bealer (1987, 1996, 1998). The main point of departure

22 / Alvin I. Goldman

is that Bealer regards intuitions as sources of a priori evidence, whereas I do not. Some may regard this as a terminological matter which should not be blown out of proportion. But Bealer seems to pin his case for (pure) rationalism in philosophical method substantially on this point, so it is worthy of close attention. I must here add the point that philosophical theorists are not interested simply in their own personal concepts associated with a given natural-language predicate. They are not merely interested in their own concepts associated with “justice”, “personal identity”, or “freedom”. They are surely interested in the concepts possessed by others as well, including non-philosophers. To obtain this information, philosophers must rely on others’ verbal reports of their intuitions, reports that must be empirically observed in order to be utilized. In this fashion, even the gathering of intuitional evidence assumes an empirical guise.20 This is not the principal point I wish to make, however, in support of the contention that empirical investigation, in particular cognitive science, has a role to play in epistemological analysis (and philosophical analysis generally). A further line of argument runs as follows.21 Epistemological analysis involves the generation and testing of hypotheses about epistemic concepts, such hypotheses as “Knowledge is justified true belief”. In principle, there are always indefinitely many hypotheses that might be advanced; and for any finite set of intuitions about cases, there are, in principle, indefinitely many hypotheses that would be compatible with those intuitions. Can cognitive science play a role in choosing among such hypotheses? Yes. Cognitive scientists investigate the form that concept representations tend to take, for example, whether they consist in sets of necessary and sufficient conditions, or weighted features, or sets of individual exemplars (see Smith and Medin 1981). Findings on this topic can bear on the relative plausibility of alternative hypotheses that epistemologists might float about epistemic concepts. For example, psychologists study how much context can affect a variety of intuitive responses to cognitive tasks. To take just one example, researchers have found that subjects’ judgements about their own level of happiness is swayed by the temporary accessibility of stimuli that shape their choice of a comparison standard. Subjects were influenced, for example, when a handicapped confederate was visible to them while they filled out the happiness report (Schwarz and Strack 1991). Such information about context effects can be relevant in assessing the relative plausibility of “contextualist” versus “invariantist” hypotheses about the meaning of “know”.22 Once it is granted that conceptual analysis has the task of laying bare the semantical features of items that are “hidden” within a cognitive system—items that are neither directly observable nor directly introspectible—it should become credible that scientific investigation is in principle relevant to the task. If anyone doubts the hiddenness of concepts (in my sense of that term), I recommend the instructive discussion by Peacocke (1998), whose term “implicit conception” corresponds to what I mean by “concept”. Peacocke points to Leibniz’s and Newton’s grappling with the notion of the limit of a series, a notion that they certainly used and understood (as well as anybody of their time) and yet

A Priori Warrant / 23

could not explicate adequately. Only in the mid-nineteenth century was a clear, unproblematic explication achieved. But even the humble word “chair” is not so trivial to explicate. Thinkers can be good at classifying cases but bad at articulating the principles guiding their classifications. Peacocke argues persuasively (to my mind) that implicit conceptions are psychologically real, subpersonal, contentful states that are useful in explaining (inter alia) classificational dispositions. He also sees the spirit of his enterprise as closely aligned with cognitive science, and cites a couple of cognitive scientists as proceeding in this spirit.23 7. Transcending the A Priori/A Posteriori Dichotomy I conclude the paper with some remarks on the need to transcend the traditional dichotomy between a priori and a posteriori knowledge. In speaking of a dichotomy I refer to the familiar fact that according to the tradition, all knowledge of warrant is either a priori or a posteriori. It cannot be both. As a matter of pure definitional stipulation, this is unproblematic. A belief either has some perceptual elements in its warranting history or it has none. If it has some, its warrant is a posteriori; if not, its warrant is a priori. This is a tenable classificational scheme, but is it very instructive? From the perspective of this paper, it is not. A significant number of people’s beliefs have a warranting history that includes both perceptual and ratiocinative processes. By calling such beliefs “empirical,” the classificational system automatically gives pride of place to the first of these components. To my mind, this is misleading. Should it be replaced with a classification system according to which beliefs with some ratiocination in their warranting history are a priori and those with no ratiocination in their warranting history are a posteriori? This would be just as inegalitarian and objectionable as the standard taxonomy. We need an epistemology that puts the two sources of warrant on a more balanced footing. The traditional terminology of “empiricism” and “rationalism” is equally misleading. As traditionally used, “rationalism” is the proper label for an epistemology which holds that some (non-analytic) beliefs are warranted a priori. But if only one such belief exists, wouldn’t the label “rationalism” inflate the significance of the a priori? What should we call an epistemology that gives roughly equal credit to perceptual and ratiocinative sources of warrant: empirico-rationalism? Unfortunately, the label “empirico-rationalism” tends to suggest that all warranted beliefs are warranted by perception, ratiocination, or a combination of the two. That, as we have seen, is false. Pure memory beliefs have only memorial warrant; they have neither perceptual nor ratiocinative warrant. Similarly, purely introspective beliefs have only introspective warrant; they too are devoid of either perceptual or ratiocinative ingredients. So it is best to reject not only the traditional options of empiricism and rationalism but even the appealing but simplistic synthesis of empirico-rationalism. Warrant is just a complex and multi-dimensional affair. Why try to force it into some neat little container or pair of containers that simply disguise its true contours? We must certainly acknowledge the rational element in

24 / Alvin I. Goldman

warrant, but this element must be assigned a suitably measured role, neither deflated nor inflated out of due proportion.24 Notes 1. For a look at contrasting forms of naturalism, both metaphysical and epistemological, see Stroud 1996. 2. For classifications of types of epistemic naturalism, see Maffie 1990 and Goldman 1994. 3. See Rey (1998) for another defense of the compatibility of naturalism and the a priori along lines similar to this one. Rey’s paper appeared when the present one was already well underway. 4. If I take this neutral position on abstractness as a condition of a priority, how can I even entertain some restriction concerning necessity? Wouldn’t necessity already commit me to abstracta? No, only if one assumes that necessity must be explicated in terms of possible worlds construed as eternal, abstract entities. If this assumption is not made, then one can entertain modal conditions on a priority while preserving metaphysical neutrality. For examples of attempts to develop non-standard approaches to modality, which do not invoke abstracta, see Field (1989), Fine (1985), and Rosen (1990). 5. Also see Plantinga 1993: 110–113. 6. Admittedly, I myself was an early proponent of the sort of trans-mental causal theory Benacerraf was addressing (Goldman 1967), but I subsequently abandoned that kind of theory (Goldman 1976). In any case, that kind of causal theory was only intended as a theory of knowledge, not of justification, which is the present topic. 7. BonJour recognizes that there are other forms of warrant (like memory and introspection) that make it impossible to identify the a priori in the simple negative fashion that Burge proposes. He proposes to capture the a priori by appeal to its deliverance of putative necessities (BonJour 1998: 8). This prospect will be discussed in the text. 8. The distinction between processes and methods—first introduced in Goldman 1986— will be explained and examined more fully later. 9. The account offered in Goldman 1992 offers further relevant details that I do not try to present here. 10. Actually, there is more than one reason offered in Goldman 1979. See that paper for elaboration. 11. Bealer (1996) rejects contingent reliability in favor of modal reliability, though his discussion concerns basic sources of evidence rather than justification, and his preferred form of modal reliability is slightly different. 12. It is controversial how a proper inductive reasoner should deal with probabilistic relations. One popular approach is that reasoners should have degrees of belief which they adjust in the light of new evidence but always so as to conform with the probability calculus. This isn’t a perfect fit with the usual presumption of theories of warrant or justification, i.e., the presumption that candidates for warrant or justifiedness are (flat-out) beliefs. So I shall not pursue the degrees-of-belief framework systematically. 13. This issue also concerns Rey (1998), who defends a reliabilist account of a priori knowledge. However, I do not understand exactly what his solution is.

A Priori Warrant / 25 14. On the no-relevant-alternatives requirement see Goldman 1976, 1986 and Dretske 1981. 15. Burge (1998) regards testimony as a potential source of a priori warrant. I do not find this general position convincing; but I am not sure that even Burge would regard the present case as an instance of a priori warrant. 16. The position of extreme domain-specificity (e.g., Manktelow and Evans 1979) says that people rely on memory of specific experiences or content-specific empirical rules to reason about logic tasks. A more moderate position, the pragmatic reasoning schemas approach, says that people use specific experiences to abstract “mid-level” rules, for example, rules that govern permissions of all sorts (Cheng and Holyoak 1985). 17. The importance of this finding is open to interpretation. For example, Rips argues that the subjects, who read a lengthy background story to explain the conditional, may have used this background information to interpret the conditional they were given. It might have “overridden” the literal meaning of this conditional sentence. Thus, relative to the interpretation they gave to the sentence, their responses may have been logically correct (Rips 1994, p. 333). 18. I do not wish to restrict all concepts to structures associated with natural-language predicates; but insofar as we are interested in philosophical analysis, that is the relevant type of concept. 19. More precisely, I would say that the process that leads from a concept to an intuition is a warranter. 20. Philosophers are not very methodologically scrupulous in this matter. They imagine that one can simply query other people about their intuitions and get a reliable reading from their answers. However, cognitive and social scientists have learned that the answers one obtains to such queries (and the intuitions that the queries generate) are partly determined by which questions are asked, how they are formulated, and the context in which they are posed. How a question is “framed” and what information is readily accessible or available to the respondent from memory can make an enormous difference to the subject’s intuitive response. (For illustrations of these ideas—though not in the precise arena of concept-instantiation—see Shafir and Tversky 1995 and Schwarz 1995.) To the extent that philosophers rely on responses to their queries as evidence, it behooves them to pay more attention to what has been learned about these matters by behavioral scientists. 21. For detailed elaboration, see Goldman and Pust 1998. 22. The contextualism/invariantism distinction is originally due to Unger (1984). For elaboration of the point in the text, see Goldman and Pust 1998. 23. Peacocke also uses the notion of an implicit conception to defend a version of rationalism; but that aspect of his discussion does not so much concern me here. 24. Thanks to Jim Tomberlin and his colleagues at California State University, Northridge, for very helpful comments and suggestions. Thanks also to Todd Stewart for excellent research assistance and commentary.

References Bealer, George (1987). “The Philosophical Limits of Scientific Essentialism,” in James Tomberlin (ed.), Philosophical Perspectives 1. Atascadero, CA: Ridgeview. Bealer, George (1996). “On the Possibility of Philosophical Knowledge,” in James Tomberlin (ed.), Philosophical Perspectives 10. Cambridge, MA: Blackwell.

26 / Alvin I. Goldman Bealer, George (1998). “Intuition and the Autonomy of Philosophy,” in Michael DePaul and William Ramsey (eds.), Rethinking Intuition: The Psychology of Intuition and Its Role in Philosophical Inquiry. Lanham, MD: Rowman and Littlefield. Benacerraf, Paul (1973). “Mathematical Truth,” Journal of Philosophy 70: 661– 679. Bigelow, John (1992). “The Doubtful A Priori,” in Philip Hanson and Bruce Hunter (eds.), Return of the A Priori. Calgary: University of Calgary Press. BonJour, Laurence (1985). The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. BonJour, Laurence (1992). “A Rationalist Manifesto,” in Philip Hanson and Bruce Hunter (eds.), Return of the A Priori. Calgary: University of Calgary Press. BonJour, Laurence (1994). “Against Naturalized Epistemology,” in Peter French, Theodore Uehling, Jr., and Howard Wettstein (eds.), Midwest Studies in Philosophy 19. Notre Dame: University of Notre Dame Press. BonJour, Laurence (1998). In Defense of Pure Reason. Cambridge: Cambridge University Press. Braine, M. D. S., Reiser, B. J., and Rumain, B. (1984). “Some Empirical Justification for a Theory of Natural Propositional Reasoning,” in G. H. Bower (ed.), Psychology of Learning and Motivation 18. New York: Academic Press. Burge, Tyler (1998). “Computer Proof, Apriori Knowledge, and Other Minds,” in James Tomberlin (ed.), Philosophical Perspectives 12. Malden, MA: Blackwell. Butchvarov, Panayot (1970). The Concept of Knowledge. Evanston: IL: Northwestern University Press. Carroll, Lewis (1895). “What the Tortoise Said to Achilles,” Mind 4: 278–280. Casullo, Albert (1988). “Revisability, Reliabilism, and A Priori Knowledge,” Philosophy and Phenomenological Research 49: 187–213. Cheng, Patricia and Holyoak, Keith (1985). “ Pragmatic Reasoning Schemas,” Cognitive Psychology 17: 391– 416. Chisholm, Roderick (1989). Theory of Knowledge, 3rd ed. Englewood Cliffs, NJ: Prentice Hall. Church, R. M. and Meck, W. H. (1984). “The Numerical Attribute of Stimuli,” in H. Roitblatt, T. G. Bever, and H. S. Terrence (eds.), Animal Cognition. Hillsdale, NJ: Lawrence Erlbaum. Cosmides, Leda (1989). “”The Logic of Social Exchange: Has Natural Selection Shaped How Humans Reason?“ Cognition 31: 187–276. Cosmides, Leda and Tooby, John (1994). “Origins of Domain Specificity: The Evolution of Functional Organization,” in Lawrence Hirschfeld and Susan Gelman (eds.), Mapping the Mind: Domain Specificity in Cognition and Culture. Cambridge: Cambridge University Press. Devitt, Michael (1996). Coming to Our Senses. Cambridge: Cambridge University Press. Dretske, Fred (1981). Knowledge and the Flow of Information. Cambridge, MA: MIT Press. Field, Hartry (1989). Realism, Mathematics and Modality. Oxford: Blackwell. Fine, Kit (1985). “ Plantinga on the Reduction of Possibilist Discourse,” in James Tomberlin and Peter van Inwagen (eds.), Alvin Plantinga. Dordrecht: Reidel. Gelman, R. and Gallistel, C. R. (1978). The Child’s Understanding of Number. Cambridge, MA: Harvard University Press. Gelman, R. and Greeno, J. G. (1989). “On the Nature of Competence: Principles for Understanding in a Domain,” in L. B. Resnick (ed.), Knowing and Learning: Issues for a Cognitive Science of Instruction. Hillsdale, NJ: Lawrence Erlbaum. Gettier, Edmund (1963). “Is Justified True Belief Knowledge?” Analysis 23: 121–123. Gibbon, J. (1981). “On the Form and Location of the Psychometric Bisection Function for Time,” Journal of Mathematical Psychology 24: 58–87. Gigerenzer, Gerd (1991). “How to Make Cognitive Illusions Disappear: Beyond ‘Heuristics and Biases’,” European Review of Social Psychology 2: 83–115. Goldman, Alvin (1967). “A Causal Theory of Knowing,” Journal of Philosophy 64: 357–372. Goldman, Alvin (1976). “Discrimination and Perceptual Knowledge,” Journal of Philosophy 73: 771–791.

A Priori Warrant / 27 Goldman, Alvin (1979). “What Is Justified Belief?” in George Pappas (ed.), Justification and Knowledge. Dordrecht: Reidel. Goldman, Alvin (1985). “The Relation between Epistemology and Psychology,” Synthese 64: 29– 68. Goldman, Alvin (1986). Epistemology and Cognition. Cambridge, MA: Harvard University Press. Goldman, Alvin (1992). “Epistemic Folkways and Scientific Epistemology,” in Liaisons: Philosophy Meets the Cognitive and Social Sciences. Cambridge, MA: MIT Press. Goldman, Alvin (1993). “The Psychology of Folk Psychology,” Behavioral and Brain Sciences 16: 15–28. Goldman, Alvin (1994). “Naturalistic Epistemology and Reliabilism,” in Peter French, Theodore Uehling, Jr., and Howard Wettstein (eds.), Midwest Studies in Philosophy 19. Notre Dame, IN: University of Notre Dame Press. Goldman, Alvin and Pust, Joel (1998). “ Philosophical Theory and Intuitional Evidence,” in Michael DePaul and William Ramsey (eds.), Rethinking Intuition: The Psychology of Intuition and Its Role in Philosophical Inquiry. Lanham, MD: Rowman and Littlefield. Harman, Gilbert (1973). Thought. Princeton: Princeton University Press. Holland, J. H., Holyoak, K. J., Nisbett, R. E., and Thagard, P. R. (1986). Induction: Processes of Inference, Learning and Discovery. Cambridge, MA: MIT Press. Johnson-Laird, Philip and Byrne, Ruth (1991). Deduction. Hillsdale, NJ: Lawrence Erlbaum. Katz, Jerrold (1998). Realistic Rationalism. Cambridge, MA: MIT Press. Kim, Jaegwon (1988). “What Is ‘Naturalized Epistemology’?” in James Tomberlin (ed.), Philosophical Perspectives 2. Atascadero, CA: Ridgeview. Kitcher, Philip (1981). “How Kant Almost Wrote ‘Two Dogmas of Empiricism’ and Why He Didn’t,” Philosophical Topics 12. Kitcher, Philip (1992). “The Naturalists Return,” Philosophical Review 101: 53–114. Kripke, Saul (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Lehrer, Keith (1990). Theory of Knowledge. Boulder, CO: Westview. Macnamara, J. (1986). A Border Dispute: The Place of Logic in Psychology. Cambridge, MA: MIT Press. Maffie, James (1990). “Recent Work on Naturalized Epistemology,” American Philosophical Quarterly 27: 281–293. Manktelow, K. I. and Evans, J. St. B. T. (1979). “Facilitation of Reasoning by Realism: Effect or Non-effect?” British Journal of Psychology 70: 477– 488. Meck, W. H. and Church, R. M. (1983). “A Mode Control Model of Counting and Timing Processes,” Journal of Experimental Psychology: Animal Behavior Processes 9: 320–334. Pastore, N. (1961). “Number Sense and ‘Counting’ Ability in the Canary,” Zeitschrift fur Tierpsychologie 18: 561–573. Peacocke, Christopher (1998). “Implicit Conceptions, Understanding, and Rationality,” in Enrique Villanueva (ed.), Philosophical Issues, 9. Plantinga, Alvin (1993). Warrant and Proper Function. New York: Oxford University Press. Putnam, Hilary (1979). “Analyticity and Apriority: Beyond Wittgenstein and Quine,” in Peter French, Theodore Uehling, Jr., and Howard Wettstein (eds.), Midwest Studies in Philosophy 4. Minneapolis: University of Minnesota Press. Quine, Willard van Orman (1961). From a Logical Point of View, 2nd ed. Cambridge, MA: Harvard University Press. Quine, Willard van Orman (1969). “Epistemology Naturalized,” in Ontological Relativity and Other Essays. New York: Columbia University Press. Reichenbach, Hans (1938). Experience and Prediction. Chicago, IL: University of Chicago Press. Rey, Georges (1998). “A Naturalistic A Priori,” Philosophical Studies 92: 25– 43. Rips, Lance (1994). The Psychology of Proof: Deductive Reasoning in Human Thinking. Cambridge, MA: MIT Press. Rips, Lance (1995). “Deduction and Cognition,” in Edward Smith and Daniel Osherson (eds.), Thinking: An Invitation to Cognitive Science, 2nd ed. Cambridge, MA: MIT Press.

28 / Alvin I. Goldman Rosen, Gideon (1990). “Modal Fictionalism,” Mind 99: 327–354. Schwarz, N. (1995). “Social Cognition: Information Accessibility and Use in Social Judgment,” in Edward Smith and Daniel Osherson (eds.), Thinking: An Invitation to Cognitive Science, 2nd ed. Cambridge, MA: MIT Press. Schwarz, N. and Strack, F. (1991). “Evaluating One’s Life: A Judgment Model of Subjective WellBeing,” in F. Strack, M. Argyle, and N. Schwarz (eds.), Subjective Well-Being: An Interdisciplinary Perspective. Oxford: Pergamon. Shafir, Eldar and Tversky, Amos (1995). “Decision Making,” in Edward Smith and Daniel Osherson (eds.), Thinking: An Invitation to Cognitive Science, 2nd ed. Cambridge, MA: MIT Press. Smith, Edward and Medin, Douglas (1981). Categories and Concepts. Cambridge, MA: Harvard University Press. Stroud, Barry (1996). “The Charm of Naturalism,” Proceedings and Addresses of the American Philosophical Association 70, 2. Newark, DE: American Philosophical Association. Tversky, Amos and Kahneman, Daniel (1974). “Judgment under Uncertainty: Heuristics and Biases,” Science 185: 1124–1131. Tversky, Amos and Kahneman, Daniel (1983). “Extensional versus Intuitive Reasoning: The Conjunction Fallacy in Probability Judgment,” Psychological Review 90: 292–315. Unger, Peter (1984). Philosophical Relativity. Minneapolis: University of Minnesota Press. Wynn, Karen (1992a). “Evidence Against Empiricist Accounts of the Origins of Numerical Knowledge,” Mind and Language 7: 315–332. Wynn, Karen (1992b). “Addition and Subtraction in Human Infants,” Nature 358: 749–750.

PHIL 13-1

example, reality cannot consist both of a system of timeless, windowless monads ... including the belief that there is a telephone on the table before me? ..... learns the truth-table method from Eileen, who explains why the method is (nec-.

110KB Sizes 1 Downloads 262 Views

Recommend Documents

131-Background Check.pdf
Evidence that the individual performed the same type of work, post. conviction, with ... Employment or character references and any other information. regarding fitness ... Applicants who refuse consent to a criminal background check will be ... the

Phil Owens.pdf
x. y. = 9 4. 39 19. −. −. ln = 4(3 ) + 3 x y. B1. M1. A1ft. forms equation of line. ft only on their gradient. (ii) x y = → = += 0.5 ln 4 3 3 9.928. y = 20 500. M1. A1. correct expression for lny. (iii) Substitutes y and rearrange for 3x. Solve

Philosophy (PHIL).pdf
... of animals,. Commented [mwh2]: GEOC and Senate approval. required (pending from last year). Page 3 of 6. Philosophy (PHIL).pdf. Philosophy (PHIL).pdf.

Adaptive Martingale Boosting - Phil Long
has other advantages besides adaptiveness: it requires polynomially fewer calls to the weak learner than the original algorithm, and it can be used with ...

Philosophy (PHIL).pdf
or 1107. Topics concern social ethics and gender, such as gender equality and the impact of gender norms on. individual freedom. Specific topics are examined in light of the intersections between gender and race,. ethnicity, class, and sexual orienta

PHIL 13-1
My principal dissatisfaction with this description of naturalistic epistemol- ogy is that no .... certain domains will qualify as warranted a priori, I want to remain neutral on the question of ..... But I do wish to register surprise that BonJour do

Print ED482778.TIF (131 pages)
Email: InternationaldevelopmentGets.org. Or visit our Web site at .... CHARLES FOWLER. Advertising Sales .... nouns, including 60% of NS speech samples.

Phil Issues final
that my laptop is not in pain, but it doesn't thereby have a reason to believe that it's .... requires knowledge of the internal world, but not the external world.9.

Fickle Consent Phil Studies - PhilPapers
my aim is to call attention to a puzzling and neglected question and hopefully to ... consent can ever justify treating him or her in a particular way. See, for .... Philosophy Conference voted 27 - 10 in favour of it being permissible for the sailor

phil collins cd.pdf
Sign in. Loading… Whoops! There was a problem loading more pages. Whoops! There was a problem previewing this document. Retrying... Download. Connect ...

man-131\mercedes-phone-activation.pdf
PDF Ebook : Verizon Wireless Phone Activation Guide. 6. ... PDF Ebook : Mercedes Benz Phone System. 15. ... PDF Ebook : Mercedes Classe A Phone System.

HCM-HPM 131.pdf
Sign in. Page. 1. /. 2. Loading… Page 1 of 2. Page 1 of 2. Page 2 of 2. "#$. %. #. &. '(). *. (). +,,. %. ,. ' !"# !$# !%# !&#. ' !"# !$# !%# !&#. - ('. ) !"# * !$# !%# + !&# ,. - . # ! #'. # ! #. ' ' . # . # ! # - . !# !# / !# ' !# !# 0 !# !# (. (

131 Carta Sanchez Martos.pdf
Whoops! There was a problem loading more pages. 131 Carta Sanchez Martos.pdf. 131 Carta Sanchez Martos.pdf. Open. Extract. Open with. Sign In.

man-131\best-accounting-textbook.pdf
man-131\best-accounting-textbook.pdf. man-131\best-accounting-textbook.pdf. Open. Extract. Open with. Sign In. Main menu. Whoops! There was a problem ...

DEPARTMENT ORDER NO. 131 .pdf
Page 3 of 60. 3. INDICE. Sinopsis ....................................................................................................................................5. Capítulo 1 .....................................................................

131 Quadratics - Rules & Graphs.pdf
GRAPH 12: y = −1x. 2 − 4x. GRAPH 13: y = −1x. 2 GRAPH 14: y = −1x. 2 + 4x. Page 1 of 1. Main menu. Displaying 131 Quadratics - Rules & Graphs.pdf. Page 1 ...

131. Mauliate ma di Ho.pdf
Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. 131. Mauliate ma di Ho.pdf. 131. Mauliate ma di Ho.pdf. Open.

Learning Halfspaces with Malicious Noise - Phil Long
Computer Science Department, University of Texas at Austin. Philip M. ... by Kearns and Li (1993) that for essentially all concept classes, it is information-theoretically im- possible ...... Journal of Machine Learning Research, 4:101–117, 2003.

Learning Halfspaces with Malicious Noise - Phil Long
Computer Science Department, University of Texas at Austin .... They also described an algorithm that fits low-degree polynomials that tolerates noise at a rate ...

PHIL-Presentation-NOV_2017.pdf
... 702-472-8556 | Email: [email protected] Website: www.phiglobal.com. PHI GROUP Business Areas of Concentration. Natural Resources. Energy. Agriculture. Healthcare. Special Situations. Company Presentation - November 2017. Page 5 of 26. PHIL-Prese

PHIL 1317: Business Ethics
QUIZZES/“BUSINESS ETHICS IN THE NEWS”:60 points possible. • EXAM 1(MIDTERM):150 points possible. • EXAM 2(FINAL):150 points possible. TOTAL POINTS:400 points possible. PHIL 1317: Business Ethics. SECTION:003. PROFESSOR:Kirsten Egerstrom. COUR

PHIL Presentation May 2018.pdf
Page 4 of 30. 5348 Vegas Drive Las Vegas, NV 89108, U.S.A. Tel: 702-475-5430; Fax: 702-472-8556 | Email: [email protected] Website: www.phiglobal.com. SCOPE OF BUSINESS. Agriculture Natural Resources. Consumer Goods Real Estate & Hospitality. Energy

PHIL Presentation May 2018.pdf
The Company intends to use acquisitions as platforms to. further acquire other targets in the same industries for its. rollup plan. Subsequently spin off acquired companies as separate. public companies. Company Presentation - May 2018. Page 5 of 30.